A Comparison of Alternative Matching Strategies for DIF Detection in Tests That Are Multidimensional

Most currently accepted approaches for identifying differentially functioning test items compare performance across groups after first matching examinees on the ability of interest. The typical basis for this matching is the total test score. Previous research indicates that when the test is not approximately unidimensional, matching using the total test score may result in an inflated Type I error rate. This study compares the results of differential item functioning (DIF) analysis with matching based on the total test score, matching based on subtest scores, or multivariate matching using multiple subtest scores. Analysis of both actual and simulated data indicate that for the dimensionally complex test examined in this study, using the total test score as the matching criterion is inappropriate. The results suggest that matching on multiple subtest scores simultaneously may be superior to using either the total test score or individual relevant subtest scores.     

Standard Errors for National Trends in International Large-Scale Assessments in the Case of Cross-National Differential Item Functioning

Standard errors computed according to the operational practices of international large-scale assessment studies such as the Programme for International Student Assessment’s (PISA) or the Trends in International Mathematics and Science Study (TIMSS) may be biased when cross-national differential item functioning (DIF) and item parameter drift are present. This bias may be somewhat reduced when cross-national DIF is correlated over study cycles, which is the case in PISA. This article reviews existing methods for calculating standard errors for national trends in international large-scale assessments and proposes a new method that takes into account the dependency of linking errors at different time points. We conducted a simulation study to compare the performance of the standard error estimators. The results showed that the newly suggested estimator outperformed the existing estimators as it estimated standard errors more accurately and efficiently across all simulated conditions. Implications for practical applications are discussed.     

Using Weighted Sum Scores to Close the Gap Between DIF Practice and Theory

We make a distinction between the operational practice of using an observed score to assess differential item functioning (DIF) and the concept of departure from measurement invariance (DMI) that conditions on a latent variable. DMI and DIF indices of effect sizes, based on the Mantel-Haenszel test of common odds ratio, converge under restricted conditions if a simple sum score is used as the matching or conditioning variable in a DIF analysis. Based on theoretical results, we demonstrate analytically that matching on a weighted sum score can significantly reduce the difference between DIF and DMI measures over what can be achieved with a simple sum score. We also examine the utility of binning methods that could facilitate potential operational use of DIF with weighted sum scores. A real data application was included to show this feasibility.     

Applying the Mantel-Haenszel Procedure to Complex Samples of Items

This Monte Carlo study examined the effect of complex sampling of items on the measurement of differential item functioning (DIF) using the Mantel-Haenszel procedure. Data were generated using a 3-parameter logistic item response theory model according to the balanced incomplete block (BIB) design used in the National Assessment of Educational Progress (NAEP). The length of each block of items and the number of DIF items in the matching variable were varied, as was the difficulty, discrimination, and presence of DIF in the studied item. Block, booklet, pooled booklet, and extra-information analyses were compared to a complete data analysis using the transformed log-odds on the delta scale. The pooled booklet approach is recommended for use when items are selected for examinees according to a BIB design. This study has implications for DIF analyses of other complex samples of items, such as computer administered testing or another complex assessment design.     

Exploring plausible causes of differential item functioning in the PISA science assessment: language,curriculum or culture

In recent years, large-scale international assessments have been increasingly used to evaluate and compare the quality of education across regions and countries. However, measurement variance between different versions of these assessments often posts threats to the validity of such cross-cultural comparisons. In this study, we investigated the cross-language, cross-cultural validity of the Programme for International Student Assessment 2006 Science assessment via three differential item functioning (DIF) analyses between the USA and Canada, Chinese Hong Kong and mainland China, and between the USA and mainland China. Furthermore, we explored three plausible causes of DIF via content analysis, namely language, curriculum and cultural differences. Our results revealed that differential curriculum coverage was the most serious cause of DIF among the three factors we investigated in this study, and differential content familiarity also contributed to DIF here. We discussed the implications of the findings for future international assessment development, and for how to best define ‘scientific literacy’ for students around the world.     

Testing Features of Graphical DIF: Application of a Regression Correction to Three Nonparametric Statistical Tests

Inspection of differential item functioning (DIF) in translated test items can be informed by graphical comparisons of item response functions (IRFs) across translated forms. Due to the many forms of DIF that can emerge in such analyses, it is important to develop statistical tests that can confirm various characteristics of DIF when present. Traditional nonparametric tests of DIF (Mantel-Haenszel, SIBTEST) are not designed to test for the presence of nonuniform or local DIF, while common probability difference (P-DIF) tests (e.g., SIBTEST) do not optimize power in testing for uniform DIF, and thus may be less useful in the context of graphical DIF analyses. In this article, modifications of three alternative nonparametric statistical tests for DIF, Fisher's χ ² test, Cochran's Z test, and Goodman's U test ( Marascuilo & Slaughter, 1981 ), are investigated for these purposes. A simulation study demonstrates the effectiveness of a regression correction procedure in improving the statistical performance of the tests when using an internal test score as the matching criterion. Simulation power and real data analyses demonstrate the unique information provided by these alternative methods compared to SIBTEST and Mantel-Haenszel in confirming various forms of DIF in translated tests.     

An Investigation of the Language Demands in Content Assessments for English Language Learners

The purpose of the present study is to examine the language characteristics of a few states' large-scale assessments of mathematics and science and investigate whether the language demands of the items are associated with the degree of differential item functioning (DIF) for English language learner (ELL) students. A total of 542 items from 11 assessments at Grades 4, 5, 7, and 8 from three states were rated for the linguistic complexity based on a developed linguistic coding scheme. The linguistic ratings were compared to each item's DIF statistics. The results yielded a stronger association between the linguistic rating and DIF statistics for ELL students in the “relatively easy” items than in the “not easy” items. Particularly, general academic vocabulary and the amount of language in an item were found to have the strongest association with the degrees of DIF, particularly for ELL students with low English language proficiency. Furthermore, the items were grouped into four bundles to closely look at the relationship between the varying degrees of language demands and ELL students' performance. Differential bundling functioning (DBF) results indicated that the exhibited DBF was more substantial as the language demands increased. By disentangling linguistic difficulty from content difficulty, the results of the study provide strong evidence of the impact of linguistic complexity on ELL students' performance on tests. The study discusses the implications for the validation of the tests and instructions for ELL students.     

Modeling Instructional Sensitivity Using a Longitudinal Multilevel Differential Item Functioning Approach

Students’ performance in assessments is commonly attributed to more or less effective teaching. This implies that students’ responses are significantly affected by instruction. However, the assumption that outcome measures indeed are instructionally sensitive is scarcely investigated empirically. In the present study, we propose a longitudinal multilevel‐differential item functioning (DIF) model to combine two existing yet independent approaches to evaluate items’ instructional sensitivity. The model permits for a more informative judgment of instructional sensitivity, allowing the distinction of global and differential sensitivity. Exemplarily, the model is applied to two empirical data sets, with classical indices (Pretest–Posttest Difference Index and posttest multilevel‐DIF) computed for comparison. Results suggest that the approach works well in the application to empirical data, and may provide important information to test developers.     

Using a Taxonomy of Differential Step Functioning to Improve the Interpretation of DIF in Polytomous Items: An Illustration

The assessment of differential item functioning (DIF) in polytomous items addresses between-group differences in measurement properties at the item level, but typically does not inform which score levels may be involved in the DIF effect. The framework of differential step functioning (DSF) addresses this issue by examining between-group differences in the measurement properties at each step underlying the polytomous response variable. The pattern of the DSF effects across the steps of the polytomous response variable can assume several different forms, and the different forms can have different implications for the sensitivity of DIF detection and the final interpretation of the causes of the DIF effect. In this article we propose a taxonomy of DSF forms, establish guidelines for using the form of DSF to help target and guide item content review and item revision, and provide procedural rules for using the frameworks of DSF and DIF in tandem to yield a comprehensive assessment of between-group measurement equivalence in polytomous items.     

Improving the Matching for DIF Analysis by Conditioning on Both Test Score and an Educational Background Variable

When tests are designed to measure dimensionally complex material, DIF analysis with matching based on the total test score may be inappropriate. Previous research has demonstrated that matching can be improved by using multiple internal or both internal and external measures to more completely account for the latent ability space. The present article extends this line of research by examining the potential to improve matching by conditioning simultaneously on test score and a categorical variable representing the educational background of the examinees. The responses of male and female examinees from a test of medical competence were analyzed using a logistic regression procedure. Results show a substantial reduction in the number of items identified as displaying significant DIF when conditioning is based on total test score and a variable representing educational background as opposed to total test score only.     

Identifying differential item functioning in multi-stage computer adaptive testing

The purpose of this study is to evaluate the performance of CATSIB (Computer Adaptive Testing-Simultaneous Item Bias Test) for detecting differential item functioning (DIF) when items in the matching and studied subtest are administered adaptively in the context of a realistic multi-stage adaptive test (MST). MST was simulated using a 4-item module in a 7-panel administration. Three independent variables, expected to affect DIF detection rates, were manipulated: item difficulty, sample size, and balanced/unbalanced design. CATSIB met the acceptable criteria, meaning that the Type I error and power rates met 5% and 80%, respectively, for the large reference/moderate focal sample and the large reference/large focal sample conditions. These results indicate that CATSIB can be used to consistently and accurately detect DIF on an MST, but only with moderate to large samples.     

DIF Detection and Effect Size Measures for Polytomously Scored Items

Data from a large-scale performance assessment ( N = 105,731) were analyzed with five differential item functioning (DIF) detection methods for polytomous items to examine the congruence among the DIF detection methods. Two different versions of the item response theory (IRT) model-based likelihood ratio test, the logistic regression likelihood ratio test, the Mantel test, and the generalized Mantel–Haenszel test were compared. Results indicated some agreement among the five DIF detection methods. Because statistical power is a function of the sample size, the DIF detection results from extremely large data sets are not practically useful. As alternatives to the DIF detection methods, four IRT model-based indices of standardized impact and four observed-score indices of standardized impact for polytomous items were obtained and compared with the R ² measures of logistic regression.     

Hierarchical Logistic Regression: Accounting for Multilevel Data in DIF Detection

The purpose of this study was to examine the performance of differential item functioning (DIF) assessment in the presence of a multilevel structure that often underlies data from large-scale testing programs. Analyses were conducted using logistic regression (LR), a popular, flexible, and effective tool for DIF detection. Data were simulated using a hierarchical framework, such as might be seen when examinees are clustered in schools, for example. Both standard and hierarchical LR (accounting for multilevel data) approaches to DIF detection were employed. Results highlight the differences in DIF detection rates when the analytic strategy matches the data structure. Specifically, when the grouping variable was within clusters, LR and HLR performed similarly in terms of Type I error control and power. However, when the grouping variable was between clusters, LR failed to maintain the nominal Type I error rate of .05. HLR was able to maintain this rate. However, power for HLR tended to be low under many conditions in the between cluster variable case.     

Decisions that make a difference in detecting differential item functioning

There are numerous statistical procedures for detecting items that function differently across subgroups of examinees that take a test or survey. However, in endeavouring to detect items that may function differentially, selection of the statistical method is only one of many important decisions. In this article, we discuss the important decisions that affect investigations of differential item functioning (DIF) such as choice of method, sample size, effect size criteria, conditioning variable, purification, DIF amplification, DIF cancellation, and research designs for evaluating DIF. Our review highlights the necessity of matching the DIF procedure to the nature of the data analysed, the need to include effect size criteria, the need to consider the direction and balance of items flagged for DIF, and the need to use replication to reduce Type I errors whenever possible. Directions for future research and practice in using DIF to enhance the validity of test scores are provided.     

Transforming SIBTEST to Account for Multilevel Data Structures

SIBTEST is a differential item functioning (DIF) detection method that is accurate and effective with small samples, in the presence of group mean differences, and for assessment of both uniform and nonuniform DIF. The presence of multilevel data with DIF detection has received increased attention. Ignoring such structure can inflate Type I error. This simulation study examines the performance of newly developed multilevel adaptations of SIBTEST in the presence of multilevel data. Data were simulated in a multilevel framework and both uniform and nonuniform DIF were assessed. Study results demonstrated that naïve SIBTEST and Crossing SIBTEST, ignoring the multilevel data structure, yield inflated Type I error rates, while certain multilevel extensions provided better error and accuracy control.     

Second language learners' performance in mathematics: Disentangling the effects of academic language features

Several studies have shown that the linguistic complexity of items in achievement tests may cause performance disadvantages for second language learners. However, the relative contributions of specific features of linguistic complexity to this disadvantage are largely unclear. Based on the theoretical concept of academic language, we used data from a state-wide test in mathematics for third graders in Berlin, Germany, to determine the interrelationships among several academic language features of test items and their relative effects on differential item functioning (DIF) against second language learners. Academic language features were significantly correlated with each other and with DIF. While we found text length, general academic vocabulary, and number of noun phrases to be unique predictors of DIF, substantial proportions of the variance in DIF were explained by confounded combinations of several academic language features. Specialised mathematical vocabulary was neither related to DIF nor to the other academic language features.     

Using Dimensionality-Based DIF Analyses to Identify and Interpret Constructs That Elicit Group Differences

In this paper I describe and illustrate the Roussos-Stout (1996) multidimensionality-based DIF analysis paradigm, with emphasis on its implication for the selection of a matching and studied subtest for DIF analyses. Standard DIF practice encourages an exploratory search for matching subtest items based on purely statistical criteria, such as a failure to display DIF. By contrast, the multidimensional DIF paradigm emphasizes a substantively-informed selection of items for both the matching and studied subtest based on the dimensions suspected of underlying the test data. Using two examples, I demonstrate that these two approaches lead to different interpretations about the occurrence of DIF in a test. It is argued that selecting a matching and studied subtest, as identified using the DIF analysis paradigm, can lead to a more informed understanding of why DIF occurs.     

Detection of Gender-Related Differential Item Functioning in a Mathematics Performance Assessment

This study used three different differential item functioning (DIF) detection proce- dures to examine the extent to which items in a mathematics performance assessment functioned differently for matched gender groups. In addition to examining the appropriateness of individual items in terms of DIF with respect to gender, an attempt was made to identify factors (e.g., content, cognitive processes, differences in ability distributions, etc.) that may be related to DIF. The QUASAR (Quantitative Under- standing: Amplifying Student Achievement and Reasoning) Cognitive Assessment Instrument (QCAI) is designed to measure students' mathematical thinking and reasoning skills and consists of open-ended items that require students to show their solution processes and provide explanations for their answers. In this study, 33 polytomously scored items, which were distributed within four test forms, were evaluated with respect to gender-related DIF. The data source was sixth- and seventh- grade student responses to each of the four test forms administrated in the spring of 1992 at all six school sites participatingin the QUASARproject. The sample consisted of 1,782 students with approximately equal numbers of female and male students. The results indicated that DIF may not be serious for 3 1 of the 33 items (94%) in the QCAI. For the two items that were detected as functioning differently for male and female students, several plausible factors for DIF were discussed. The results from the secondary analyses, which removed the mutual influence of the two items, indicated that DIF in one item, PPPl, which favored female students rather than their matched male students, was of particular concern. These secondary analyses suggest that the detection of DIF in the other item in the original analysis may have been due to the influence of Item PPPl because they were both in the same test form.     

Differential Performance by English Language Learners on an Inquiry-Based Science Assessment

The performance of English language learners (ELLs) has been a concern given the rapidly changing demographics in US K-12 education. This study aimed to examine whether students' English language status has an impact on their inquiry science performance. Differential item functioning (DIF) analysis was conducted with regard to ELL status on an inquiry-based science assessment, using a multifaceted Rasch DIF model. A total of 1,396 seventh- and eighth-grade students took the science test, including 313 ELL students. The results showed that, overall, non-ELLs significantly outperformed ELLs. Of the four items that showed DIF, three favored non-ELLs while one favored ELLs. The item that favored ELLs provided a graphic representation of a science concept within a family context. There is some evidence that constructed-response items may help ELLs articulate scientific reasoning using their own words. Assessment developers and teachers should pay attention to the possible interaction between linguistic challenges and science content when designing assessment for and providing instruction to ELLs.